Abstract: As for the poor generalization of the traditional diagnosis method of rolling bearing, this paper proposes a fault diagnosis method of rolling bearing based on time-frequency representations and CNN. Firstly, the vibration signals are converted into short-time Fourier transform to construct the time-frequency feature maps. Then, the time-frequency representations of the training data are used as the inputs of the Convolutional Neural Network to train the network model. Finally, the time-frequency representations of the test data are input to the network model so as to identify the fault situation of rolling bearing. Multiple sets of validation experiments were carried out with the open dataset at Case Western Reserve University. The results show that this method can effectively determine whether the bearing is faulty and identify the classification of fault, and the accuracy rate can reach more than 97.63%.